lack of consensus on coercion

CSGnet Archive CSG1998
1

[From Bruce Nevin (980519.0901 EDT)]

Rick Marken (980518.1600)--

If we can't get consensus from modelers then we do, indeed, have a
problem.

Has there been any actual model of coercion or of RTP processes built and
tested yet? That might account for the lack of consensus.

Not that the discussion of coercion is pointless. You have to identify a
phenomenon with some precision before you can model it. The process of
modelling can change our ideas about the phenomenon. We have to be careful
then about ignoring associated phenomena because they don't arise from the
model. Naturalistic description is a necessary starting point and an
ongoing touchstone.

I don't think we have a problem, it's just that we're at an early stage of
getting our heads around the phenomenon of coercion. We don't have a model.
Rick, your demo does not control my actions. It does not stop me from
clicking the Down button (if I want a lower value). It does not require me
to perform certain actions, by comparing them to a reference perception of
my actions. It does not have any perception of my actions. It is an
interesting step in the "different worlds" issue, but doesn't say anything
about coercion, other than the participant's experience of frustration.

And not that the discussion of coercion in RTP is pointless. What we have
is statements of principle and guidelines for practice. We don't have a
model. We don't have the prerequisite for modelling, naturalistic
description of actual process in the field. This discussion opens up
aspects of such a description, aspects that perhaps are not reflected in
the statements of principle and guidelines for practice. Just as the
results of behaviorist experiments depend upon covert coercion (prior
deprivation defines what is rewarding), some background coercion provides a
structure that enables RTP to work. But this is not a Bad Thing. That
coercive structure is inherent in parenting, and in the teacher role as an
extension of parenting, and necessarily so because as Bill points out
(980519.0143 MDT) where children aren't yet capable of controlling their
parents must control for them.

One thing missing from this account is how a child reaches out to new
growth (this may be perceived as rebellion or incompetence etc.) and
retreats to familiar areas where control is secure (this may be perceived
as being good or as regression etc.) in a cyclical way. How much to let
them take their own falls and learn, and how much to save them from injury,
is a continual judgment call. Even more important is the child learning to
recognize this process and participate in it intelligently, since it does
go on throughout life, and as adults we must learn to push boundaries or
retreat and assimilate when each is appropriate according to our own inner
tides. This is akin to learning to learn. Getting the wisdom to be wise.
The example of wise parents and wise teachers is important for this. RTP
could provide a good structure for this.

  Bruce Nevin

2

[From Rick Marken (980519.1020)]

Bruce Nevin (980519.0901 EDT)--

Has there been any actual model of coercion or of RTP processes built and
tested yet?

The "Different Worlds" demo

http://home.earthlink.net/~rmarken/ControlDemo/Worlds.html

is a model of coercion. When an opposing control system is present
and your aim is to control the sum of the lower two numbers, the
opposing control system, which is also controlling the sum of these
two numbers, is coercing this sum to the value _it_ wants. This is
coercion because 1) the opposing control system is controlling
some aspect of _your_ behavior (the sum of the two lower numbers)
and 2) doing this controlling _sucessfully_ because it can contribute
more
to this sum, and do it faster, than you can.

Rick, your demo does not control my actions.

No. It controls one of your controlled variables. It's analogous to the
following situation: You want the door open 20 degrees, I want it
closed. We are both pushing on the door (from opposite sides) but I am
much stronger and faster than you so I am able to keep the door where
I want it. So I have coerced one of your behaviors (a variable you are
controlling -- door angle) into the state _I_ want it in (closed)
without
controlling your actions (the force you exert on the door).

And not that the discussion of coercion in RTP is pointless.

The discussion has the point of showing how the PCT model applies
to the behavior of the teachers and administrators in the RTP program
who are charged with controlling the behavior of disruptive students.
The RTP practitioners are charged with controlling a perception of the
"disruptiveness" of the behavior of the kids in the classroom. If a
kid's
behavior differs from the reference (no disruption in class) then, after
the second offense, the kid is removed from class. This is control
of behavior. It's _very_ simple to model. It is an obvious application
of the PCT model. It is what is typically called "coercion".

We don't have a model.

Sure we do. My simple little control model in the "Different Worlds"
demo
is a model of a coercer. A coercer is just a control system that can
successfully control some aspect of the behavior of another control
system.
My little model coercer doesn't "know" it is coercing; it's just
controlling.
We (the observer) know it's coercing because we can see that the
variable
it is controlling is an aspect of the behavior of another control system
(us).

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

3

[From Richard Kennaway (980519.1843 BST)]

Rick Marken (980519.1020):

It controls one of your controlled variables.

...

A coercer is just a control system that can
successfully control some aspect of the behavior of another control
system.

What does it mean to say that a controlled variable is an "aspect" of the
behaviour of the control system that is (attempting to) control it? Am I
coercing someone every time I succeed in controlling a variable that they
are trying to control at a different value?

In the RTP the teachers, by ejecting disruptive students (using whatever
force is necessary), are controlling a part of those students' behaviour.
In your demo, what the supposedly coercive system is controlling is rather
a long way removed from the subject's behaviour. The word "aspect" looks
to me like an attempt to create a category to contain them both, so as to
maintain the claim that the latter is a model of the coercion present in
the former. I don't think it works.

-- Richard Kennaway, jrk@sys.uea.ac.uk, http://www.sys.uea.ac.uk/~jrk/
   School of Information Systems, Univ. of East Anglia, Norwich, U.K.

4

[From Bruce Gregory (980519.1420 EDT)]

Rick Marken (980519.1020)

Bruce Nevin (980519.0901 EDT)--

> Rick, your demo does not control my actions.

No. It controls one of your controlled variables. It's analogous to the
following situation: You want the door open 20 degrees, I want it
closed. We are both pushing on the door (from opposite sides) but I am
much stronger and faster than you so I am able to keep the door where
I want it. So I have coerced one of your behaviors (a variable you are
controlling -- door angle) into the state _I_ want it in (closed)
without
controlling your actions (the force you exert on the door).

PCT does use words in _strange_ ways. If I open the window, I am apparently
coercing the behavior of the thermostat. Sounds a little weird, but if
that's the way you want to use the word, who am I to argue? It is helpful,
however, to point out that this usage bears little resemblance to the way
the word is used in the real world.

Best Offer

5

[From Rick Marken (980519.1540)]

Richard Kennaway (980519.1843 BST)--

What does it mean to say that a controlled variable is an "aspect"
of the behaviour of the control system that is (attempting to)
control it?

         r
         >
         v
      --|C|--
    p> >e

···

      > system

---------------------
     > > environment
     > v
d -->q<------o

I basically mean q (the controlled variable) and o (the
output variable). These two variables are aspects of the
behavior of the control system (they are part of the
control loop) that are _accessible_ to a would-be coercer.
Other aspects of the behavior of the control system are
e, p and r. I don't consider d (disturbance variable) an
aspect of the behavior of the control system because it
is independent of the operation of the control loop _and_
it is outside the system (r is also independent of the operation
of the control loop but it is _inside_ the system so variations
in r count as an aspect of the behavior of the control system).

Am I coercing someone every time I succeed in controlling a
variable that they are trying to control at a different value?

I'd say "yes".

In your demo, what the supposedly coercive system is controlling
is rather a long way removed from the subject's behaviour.

Behavior is the control of perception. In the demo the coercive
system is controlling the perception controlled by the subject.
So I'd say that the coercive system is controlling the subject's
behavior rather directly -- at least, when the subject is
controlling the sum of the lower two numbers.

The word "aspect" looks to me like an attempt to create a category
to contain them both

Sorry. I don't really know how to say it better. The problem is
that the word "behavior", as conventionally used, is ambiguous
from a PCT perspective; it can refer either to a controlled
perception (like "book at position X"), the outputs (actions)
that are used to control the perception (like "the muscle forces
that bring the book to position X") _or_ to irrelevant side effects
of these actions (like "the arm moves upwards").

Best

Rick
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

6

[From Tim Carey (980520.1035)]

[From Rick Marken (980519.1540)]

> Am I coercing someone every time I succeed in controlling a
> variable that they are trying to control at a different value?

I'd say "yes".

So the flip side of this would be that if you succeed in controlling a
variable to the same value that someone else wants it controlled then it's
_not_ coercion.

Cheers,

Tim

7

[From Bill Powers (980520.0745 MDT)]

Tim Carey (980520.1035)--

[From Rick Marken (980519.1540)]

Tim:

> Am I coercing someone every time I succeed in controlling a
> variable that they are trying to control at a different value?

Rick:

I'd say "yes".

Tim:

So the flip side of this would be that if you succeed in controlling a
variable to the same value that someone else wants it controlled then it's
_not_ coercion.

No, Tim. You're just doggedly insisting that coercion exists only when
there is an active force being applied to the coercee. Apparently, that's
what you're going to insist on no matter what anyone else says, but at
least do us the courtesy of not putting words in our mouths. Neither Rick
nor I would say that there's no coercion under the conditions you describe.
The only way for no coercion to exist would be for the first person not
even to try to control the variable that someone else is controlling, or
not to be stronger than the other person or have greater resources.

Best,

Bill P.

8

[From Bill Powers (980520.0605 MDT)]

Richard Kennaway (980519.1843 BST)--

Rick Marken (980519.1020):

A coercer is just a control system that can
successfully control some aspect of the behavior of another control
system.

Richard:

What does it mean to say that a controlled variable is an "aspect" of the
behaviour of the control system that is (attempting to) control it?

[See below]

Richard:

Am I
coercing someone every time I succeed in controlling a variable that they
are trying to control at a different value?

Bill:
Yes, by my definition and Rick's, and maybe yours if you'll accept it. In
order for you to succeed at controlling a variable someone else is trying
to control at a different value, you must be strong enough, or have
resources enough, to overcome the maximum control action that the other
person can produce. Otherwise, the controlled variable will end up
somewhere between the value you intend and the value the other person
intends, and your control will not succeed.

Richard:

In the RTP the teachers, by ejecting disruptive students (using whatever
force is necessary), are controlling a part of those students' behaviour.

Bill:
Yes, their position in space.

Richard:

In your demo, what the supposedly coercive system is controlling is rather
a long way removed from the subject's behaviour. The word "aspect" looks
to me like an attempt to create a category to contain them both, so as to
maintain the claim that the latter is a model of the coercion present in
the former.

Bill:
An "aspect" is one perceived dimension of the overall pattern of activity
that an observer would informally call the behavior of an organism. It is
defined by the perceptual function that receives inputs from the
environment (which can mean lower-level perceptions) and computes a
particular function of them, the value of which is represented as a
perceptual signal. There are as many "aspects" of a given environment as
there are ways of constructing input functions to represent it.

The generalization Rick is using simply acknowledges that behavior is
organized at many levels, and conflicts can occur at many levels. In a
poker game, for example, there can be a conflict over who will win the pot.
The betting and the raising are opposing actions, although the physical
force and energy involved are trivial. There is a coercive strategy of
betting in which one player make a bet so large that none of the other
players can match it. If no credit is allowed (no "going light"), the other
players are automatically out when they can't match or raise the bet. At
that that point the coercive bettor wins, without showing his hand. To
prevent use of this strategy, a rule can be adopted that limits the size of
a raise and the number of raises. This at least prevents the player with a
very large pile of money from "buying the pot".

Richard:
I don't think it works.

Bill:
It does if you don't limit your definition of behavior so it includes only
first-order controlled variables.

Best,

Bill P.